Distributed load management supporting power injection and reactive power balancing

In recent years active control over selected electric loads has become increasingly important to reduce the peak demand and allow integration of intermittent renewable generation into the power grid. Two groups of loads are especially interesting in this regard due to their expected numbers and large freedom in scheduling: electric vehicles and so called thermostatically controlled loads, such as refrigerators, hot water heaters, or air conditioners. These electric loads allow their power consumption to be adapted depending on the needs of the distribution grid with a minimal impact on the customers. We consider two types of load power control abilities: binary and continuously controllable power. In this project we propose a load management scheme to deal with such electric loads. The load management scheme allows the usage of two algorithms: one for binary and one for continuously controllable loads. The proposed load management scheme relies on broadcast signals that are sent by a central management unit to all the agents connected. This means that the communication load is low and that simultaneous management over different load types is possible, due to the identical set up. Further, as there is no data transmitted from the controllable loads to the central management unit, there are no data protection or privacy issues present. For loads participating with binary controllable power consumption, we propose a binary automaton algorithm that uses stochastic decisions made by the agents to govern the power consumption. This algorithm’s behaviour is analysed and shows promising behaviour in simulations. The algorithm we propose for handling continuously controllable loads is the additive increase multiplicative decrease (AIMD) algorithm. This algorithm is commonly used for congestion control in communications networks and has shown to be very flexible and reliable. Its behaviour has been investigated in detail. While there are some adaptations needed to apply this algorithm in a load management case, we can apply many of the existing results found for the AIMD algorithm as it is applied in congestion control. We extend the analysis where necessary for our case.